JPH076157A - Cooperative problem solving system for assignment problem - Google Patents

Abstract

PURPOSE:To facilitate the derivation of the solution by dividing a large scale problem into subproblems, finding out local solutions and unifying local solutions found out in respective agents. CONSTITUTION:Each agent stores information 7 related to a resource such as a network and knowledge 8 for solving a problem in its data base. Request distributing processing 1001 distributes respective assignment requests to respective agents. assignment resource information storing processing 1003 stores information related to a resource assigned by a local solution obtained by a local solution calculating means 1002 in an information data base 1007 related to the assigned resource. Resource competition checking processing 1004 judges the existence of competition of a local solution independently calculated in each agent of other agents. When the competition exists, local solution information exchanging processing 1005 exchanges information related to local solutions between plural agents under competition. Finally, end judging processing 1008 ends the assignment of resources when all agents are turned to a message waiting state.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an allocation problem-oriented cooperative problem solving system for allocating a limited amount of resources such as route setting on a network, LSI wiring design, facility layout, etc. on demand, and more particularly, In the present invention, when the resource allocation request does not have locality, when the resource amount to be allocated, the number of allocation requests are large, or when the request is dynamically changed, these conditions are The present invention relates to a collaborative problem solving system for assignment problems that can allocate resources in a practical time even when combined.

[0002]

2. Description of the Related Art FIG. 11 is a diagram for explaining a conventional resource allocation problem.

A problem of allocating resources such as a network to a request such as route setting, which is a typical conventional allocation problem, will be described with reference to FIG.

FIG. 1A shows the network as an undirected graph composed of nodes and edges 2 having a capacity of a positive integer. In the figure (a), the numbers (20, 10) indicate the capacitance of each edge. In the network of FIG. 9A, as shown in FIG. 9B, the problem of setting a route having a capacity that satisfies the resource requirement 3 between nodes is defined as a network resource allocation problem. In the figure (b), the numeral (10,
5) indicates the required amount of resources. In the figure, node 11
The request amount 10 is allocated between the node 12 and the node 12 and the node 13, and the request amount 5 is allocated between the node 13 and the node 14 and between the node 11 and the node 13. FIG. 14C shows a result 4 obtained by allocating the request 3 of FIG. 11B to the resource (edge 2) of FIG. 11A by the method described below.

FIG. 12 is a diagram for explaining a network resource allocation problem in which resources such as a network are allocated to requests such as route setting. In the following description, the problem that the use of resources requires some cost is targeted, but the problem of not considering the cost can be dealt with by making all costs the same.

FIG. 1A shows the network as an undirected graph composed of a node 2 and a link 1 having a capacity of a positive integer. The resource is link 1 of the undirected graph, and the link capacities are all 1. The numerical value added to the link 1 indicates the cost 3 per usage amount of the link 1.

In the network shown in FIG. 1A, a route is allocated so that the capacity of each link 1 is not exceeded in response to the route setting request 4 (plurality) between nodes as shown in FIG. Is the network resource allocation problem.
Here, as a problem considering cost, consider a network resource allocation problem in which the cost 3 required to use the link 1 is minimum (maximum).

FIG. 3C shows the allocation result 5 for the request 4 in FIG. The capacity 1 is allocated to the links between the node a and the node b, the node b and the node e, the node e and the node h, and the node h and the node i, and the cost of using the link of the entire allocation route is 5.

Since such a problem can be reduced to a linear programming problem, the simplex method and the Carmarker method (literature: "Karmarker, N .:
A New Polynomial-Time Algorithm for Linear Progra
mming, Combinatorica, No.4, 1984 ”) has been proposed. Linear programming is
This is a problem of maximizing or minimizing the objective function under the constraint condition. From the linearity of the objective function and the constraint condition, the optimum value exists at the end point of the constraint region, so it is optimal to check all the end points of the constraint region. Can find a solution.

The simplex method takes advantage of this property, but when the scale of the problem is large (the number of variables is large), the number of end points increases exponentially, which makes it difficult to apply.

The Carmarker method searches for an optimum solution based on the steepest descent direction (search direction vector) of the objective function, and is an interior point method that approaches the end point of the optimum solution from inside the restricted region. Since it is not directly affected by the number, the optimal solution can be obtained in polynomial order.

In such an analytical algorithm, one problem solver solves a problem in the entire network, so that information on the entire network can be viewed, and it is easy to understand the relationship between requests.

FIG. 13 is a diagram for explaining the area division type resource allocation problem.

The example of the figure shows a method of dividing the problem space on the premise of regionality (called regional division).

In the example shown in the figure, a network with n × n nodes uniformly connected in a grid pattern is assumed, and route setting requests are given from the upper left to the lower right and from the lower left to the upper right of the network. In the area division, this network is divided into two, a and b, and agents A and B take charge of resource allocation in the left and right areas. Agents A and B seek local solutions that satisfy their respective routing requirements within a given area.

When the allocation results by the agents A and B in FIG. 13 are checked, when the area division is performed and the edge 2 allocated at the boundary surface is different (a route is set for one route setting request, In the case of discontinuity at the boundary of regions), a negotiation C for exchanging problem solving information between agents A and B occurs.

[0017]

However, it is difficult for the algorithm used as the OR method as described above to obtain the directionality of the effective search direction vector and the step size (update width of the variable value).

Further, since such a problem is originally NP perfect in the theory of complexity of calculation, when the scale is large (the number of nodes is several tens or more), the optimum solution should be derived in a practical time. It is difficult. Furthermore, it becomes more difficult in situations where the demand changes dynamically,
On the other hand, the need for finding an optimal solution is not very high.

In such a case, the problem is divided into several parts, and each partial problem is stored in a database with information about resources such as a network and knowledge for solving the problem, and the information and knowledge are used. A method has been proposed in which an agent that solves a problem is in charge and a solution is obtained by cooperation between the agents.

Particularly, in the area division type resource allocation problem shown in FIG. 13, the possibility of mismatching of the boundary surfaces increases and the probability of negotiation C increases, so that the problem space cannot always be divided conveniently. Also, when dividing into regions, it is difficult to see the state of the entire network.

The present invention has been made in view of the above points, and is different from the conventional method in which a local solution is obtained for each area and then resource allocation is adjusted between the local solutions to obtain a final solution.
Like the task division type resource allocation problem, the problem is divided by focusing on individual requests for resource allocation (task division), and the agent resources are coordinated by agents arranged for each one or more requests for resource allocation. It is an object of the present invention to provide a cooperative problem solving system for assignment problems that can reduce the occurrence of conflicts among competing agents and can respond to dynamically changing requests.

[0022]

FIG. 1 is a block diagram showing the principle of the present invention.

The allocation problem-oriented cooperative problem solving system of the present invention is a system for solving an allocation problem in which a plurality of limited resources are allocated to a plurality of allocation requests, and an agent which executes an allocation process for each allocation request 100. The request distribution means 110 for determining and the knowledge for the agent 200 to obtain the optimum solution for the resource-related information and the allocation request 100 for solving the problem and the resource conflict that has occurred between the agent's solution and the knowledge. Problem solving knowledge 300 including knowledge for solving, resource information 400 regarding resources to be allocated, and each agent 200
Using the problem solving knowledge 300 and the resource information 400,
A local solution calculating unit 220 that obtains a local solution that satisfies the allocation request distributed by the request distributing unit 110, a resource information storage unit 230 that stores information about resources allocated by each agent by the local solution calculating unit 220, and resource information. The storage means 230 is referred to, and a resource allocation solution 50 is generated according to the constraint on the resource of the local solution obtained by each agent among all the agents and the state of competition with other agents.
Allocation solving means 240 for determining 0.

In addition, the allocation solving means 240 of the present invention uses the competition occurrence judging means 241 for checking the presence or absence of resource competition between the local solutions calculated by the local solution calculating means 220 by each agent, and the competition occurrence judging means 241 for competing. In the case where there is a resource, the resources are extracted from the local solution information exchanging means 242 that spontaneously synchronizes the agents that are in contention with each other and exchanges information about the local solution, and the information exchanged by the local solution information exchanging means 242. Again to resolve the conflict,
Request determining means 243 for determining a request for searching for a local solution
, And the agents that are competing for resources resolve the conflict independently.

Further, the local solution information exchange means 242 of the present invention.
Includes a competing resource information exchanging means 2421 for exchanging information on resources for which resource competition has occurred so far, and a resource removing means 2422 for removing resources from the objects used for allocation.

[0026]

The present invention divides the problem into sub-problems from the aspect of the request, and distributes the requests such as the route setting generated for all the target resources to the agents in one or a plurality of groups. Each agent uniquely solves a partial problem (local solution), that is, the agent seeks a solution that satisfies the assigned requirements regardless of the requirements assigned to other agents. Is a system that seeks a solution as. The present invention solves a problem for each agent using information and knowledge in a database,
Divide a large-scale problem into subproblems and obtain a local solution,
We try to facilitate the derivation of the solution by integrating the local solutions found for each agent.

At this time, if there is a conflict when allocating resources with another agent, the agents involved are exchanging information about the local solution, and this information is used to prevent subsequent conflict occurrence. Determine the demand for resource reallocation to reduce.

FIG. 2 is a flow chart for explaining the operation of the present invention. First, a plurality of requests are first equally distributed to each agent (step 200). Next, each agent independently sets a route for all distributed unallocated requests. In this case, the route setting results of other agents are not considered (step 201). When the route setting of all agents is completed, it is checked whether resource conflict has occurred. When a resource conflict is found (step 203), a route (request) bypassed between the agents causing the resource conflict to resolve the conflict is determined by negotiation. After the route to be bypassed is determined, the start point of the conflicting point is deleted and the process returns to step 201 (step 204). If there is no resource conflict (step 203), the route setting of all requests is completed (step 205).

Further, the agent in charge of the request that needs re-allocation removes the competing resource from the resource used for allocation, and then again obtains a local solution for each agent, while the problem solving status of each agent is determined. Check to determine the end of calculation.

As a result, each agent arranged for each subproblem locally solves the problem, and the cooperation of these agents facilitates the solution of a large-scale and non-local problem. It is possible to reduce the occurrence of resource competition or solve the problem even if resource competition occurs while performing problem solving in an asynchronous parallel manner.

[0031]

Embodiments of the present invention will now be described in detail with reference to the drawings.

According to the present invention, a problem is divided into several subproblems from the aspect of request, an agent for solving a problem for a request such as route setting is arranged for each subproblem, and each agent has a partial subproblem. It is a system that solves a problem and a plurality of agents collaborate with each other to obtain a local solution, which is the result of solving a partial problem.

FIG. 3 is a diagram for explaining the task division type resource allocation problem used in the present invention. The allocation target resource is edge 2 and the edge capacity is all 1. The “request” is a route setting request from the upper left node to the upper right node. A route setting request from the lower left node to the upper right node. In the figure, tasks are divided so that one agent A and one agent B are assigned to each route setting request. As a first step, each agent A, B independently allocates and obtains an initial solution. Figure 3
Shows an example of the result of obtaining this initial solution. In this example, one edge is used by two agents, and it can be seen that negotiation between the agents is required for the edge location.

FIG. 4 shows the system configuration of an embodiment of the present invention. As shown in the figure, the agent 5 has information 7 about resources such as networks (hereinafter referred to as resource information) and knowledge 8 for solving problems in the database.
(Hereinafter referred to as "knowledge") and the like, and includes a problem solving engine 6 that solves a problem using the resource information 7 and the knowledge 8. The problem solving engine 6 performs a process including the following two phases.

Phase 1: Within the agent, a local solution that satisfies the request placed in the agent is obtained.

Phase 2: Information on partial problem solving is exchanged between a plurality of agents so as to satisfy the constraint that the resource capacity is not exceeded during the local solution obtained by each agent between the agents (referred to as negotiation). ), While working together to solve.

Phase 1 is a process that each agent performs asynchronously independently, and phase 2 shows a process between arbitrary plural agents, but each phase is not completely cut off.

FIG. 5 is a diagram for explaining the outline of the operation of the embodiment of the present invention, and FIG. 6 is a diagram for explaining the problem solving by each agent of the embodiment of the present invention. .

First, as a phase 1, the request distribution processing 10
01 distributes each allocation request to the agents. Moreover, a plurality of allocation requests may be allocated to one agent (step 1).

Next, each agent, in the local solution calculation processing 1002, obtains a local solution that satisfies the distributed requirements by using the resource information 7 and the problem solving knowledge 8 (step 2). Here, the process of step 2 is performed for each agent to which the allocation request is distributed. Further, it is assumed that the resource information 7 and the problem solving knowledge 8 are stored in the database.

The local solution calculation process will be described in detail below. Each agent independently reserves a set of edges that satisfy the route setting request. Which route is selected by each agent may be determined by its own judgment (for example, the shortest route, the minimum cost route, etc.), and is not limited.
Considering that each agent reserves its own resources, conflicts of interest between agents can occur. Therefore, in order to reduce the probability that negotiation is required in phase 2, when a local solution is obtained by some or all agents, a strategy of not selecting a route that is likely to become a bottleneck is effective.

Here, as one method for calculating the local solution, a method considering the link occupancy will be shown.

<< Local Solution Calculation Considering Link Occupancy Ratio >> Let the link occupancy ratio be [(amount used by agent for allocation) / (link capacity)]. In the route search at the time of local solution calculation, the link occupancy rate is calculated every time a link is used for allocation, and if the preset threshold is exceeded, resource conflict check is performed in advance, and only links that do not become resource conflict To find the local solution.

If the threshold value is set to 0, resource competition is always checked every time a link is made, so resource competition does not occur and negotiation can be prevented, but the overhead of resource competition check increases. On the other hand, if the threshold value is set to 1, no resource conflict is checked when the link is used, so that no overhead occurs, but resource conflict occurs later and the possibility of negotiation increases. By appropriately setting this threshold value from the relationship between the cost of the resource conflict check and the cost of the negotiation, the problem can be solved efficiently.

Next, the allocated resource information storage processing 1003 in phase 2 stores the information about the resources allocated by the local solution obtained by the local solution calculation processing 1002 in the allocated resource information database 1007. Here, it is assumed that the information database 1007 regarding the allocated resources stores information about the resources to which all the agents have allocated resources based on the local solution.

Next, the resource conflict check processing 1004
Determine the competition with other agents for the local solution calculated by the agent. As described above, when an agent locally obtains a local solution, resource competition with a local solution of another agent may occur. Therefore, each time an agent obtains a local solution by the local solution calculation processing 1002, an allocated resource is allocated. The resources used for allocation are registered in the database 1007 by the information storage processing 1003. Then, in the resource conflict check processing 1004, each agent independently refers to the resource information of the database 1007 to check the presence / absence of resource conflict for its own local solution.

If there is a conflict due to the resource conflict check processing 1004, the local solution information exchange processing 1005 exchanges information regarding the local solution between the agents having the resource conflict (step 3). If there is a resource conflict, a negotiation message is transmitted to the agent in the resource conflict in the information exchange (step 4). On the other hand, if there is no resource conflict, it is checked whether a negotiation message has arrived from another agent (step 5).

In the local elimination information exchange processing 1005,
If the negotiation message has arrived, the process proceeds to the negotiation of the resource conflict resolution processing 1006 (step 6), and if it has not arrived, it is checked whether there is an unprocessed request (unprocessed request check) (step 7).

If there is an unprocessed request, the process returns to the local solution calculation process 1002 again, and if there is no unprocessed request, the process moves to the message check. In the message check, the reception status of the negotiation message and the end message is checked, and if the message has not arrived, the message waits.

Finally, the termination judgment processing 1008 monitors the states of all the agents, and when all the agents wait for a message, sends a termination message indicating the end of calculation to all the agents to terminate the resource allocation ( Step 8).

Next, here, the resource conflict resolution processing 1006
The process of resolving the resource conflict by the negotiation will be described.

<< Resolution of Resource Conflict by Negotiation >> Negotiation is a process of exchanging information between agents that are in a resource conflict to eliminate a resource conflict.

FIG. 7 is a sequence chart showing the procedure of negotiation according to the embodiment of the present invention.

The figure shows a situation in which resource competition has occurred among agents A, B and C. The agent A, which has found the resource conflict, sends a negotiation message to the other agents B and C that are in the resource conflict (step 10).
0). Agents B and C receiving the negotiation message
Information regarding the local solution is transmitted to the agent A as a local solution message (step 101).

The agent A uses these pieces of information to determine a route to be bypassed in order to resolve the resource conflict based on a method described later. After this is determined, the fact is transmitted to the agents B and C as a determination message (step 102).

The agent C, which has received the detour need determination message, re-searches the detour route as a new local solution, excluding the link in which the resource conflict has occurred. The agents A and B, which do not need detour, continue the state before negotiation. In the example of the figure, the agent A is in the local solution search continuation state, and the agent B is in the message waiting state.

In this embodiment, the agent C, which has received the message indicating that the detour is necessary, enters the re-search for the local solution (detour route search) according to the message, but a procedure for rejecting this may be considered. In this case, information is repeatedly exchanged with the agent A until they are satisfied with each other.

Also, in the information exchange at the time of negotiation, the information of the links that have been eliminated due to the competition so far is also exchanged, and the search is also performed by excluding these links in the re-search process. It is possible to prevent the resource competition from occurring in the link and reduce the resource competition.

FIG. 8 is a diagram for explaining a problem applied to the network resource allocation problem of the embodiment of the present invention.

FIG. 7A shows the status of network resources, and FIG. 7B shows a route setting request. In the network resources of FIG. 5A, the numerical value between nodes indicates the capacity of edge 2. The numbers in FIG. 7B show the required amount.

In the present embodiment, a solution method when two resource allocation requests 3 are given as shown in FIG.

FIG. 8 is a diagram for explaining the partial problem solution by each agent applied to the network resource allocation problem of one embodiment of the present invention. In the figure, each agent independently obtains its local solution. The figure (a) has shown the allocation result 9 of one agent A, and the numerical value (10) is the allocation amount. The figure (b) shows the allocation result 10 of the other agent B, and the numerical value (10) is the allocation amount. In this example, agent B is making a detour.

In the figure, each agent calculates a solution satisfying the capacity constraint of each edge.
The criteria for this allocation may be the individual judgment of each agent, such as the shortest route search by the Dijkstra method.

In this embodiment, an edge with the shortest path and a loose capacity constraint (capacity as large as possible) is selected.

FIG. 10 is a diagram for explaining the negotiation between agents applied to the network resource allocation problem of one embodiment of the present invention.

FIG. 9A shows a state in which local solutions are matched, and FIG. 9B shows a state in which capacity constraint violation is avoided by negotiation between agents. In the example of the figure, the agent B makes a detour.

Next, the case where the local solutions are matched between the agents will be described. If there are two agents allocating to the same edge, it is checked whether the sum of the request amounts allocated by each agent exceeds the total capacity of the edge. if,
If it exceeds, avoid it by negotiation between agents. Negotiation methods include low cost for detour candidate search and small reduction in utility of each agent due to avoidance.

Examples of the negotiation method between the agents include the following methods.

Avoid by using the priority of each agent. It should be noted that the priority order may be determined statically in advance even if all the agents are statically determined in advance, or the priority order may be dynamically determined when the capacity constraint violation occurs.

An agent having a low priority order finds an alternative route by searching the route again, assuming that there is no edge in which the capacity constraint is violated.

When an agent with a low priority cannot obtain an alternative route, an agent with a high priority determines that there is no edge in which the capacity constraint is violated, and obtains an alternative route by searching the route again. .

Each agent brings in alternative candidates and makes a judgment. Each agent finds an alternative route by searching the route again assuming that there is no edge in which the capacity constraint violation has occurred.

The alternatives obtained by the agents are compared with each other, and the alternative to be selected is determined based on the cost of avoidance.

Next, a method of deciding a route to be bypassed at the time of negotiation will be described.

As a first method, there is a method of making a decision based on the cost of the current route. This method compares the sizes of the current route costs and detours the routes in descending order of cost.

As a result, a route with a high cost is a route using a link with a small capacity, a route with a large number of via links, or a route in which both are combined.
Since resource competition is likely to occur in a link having a small capacity, moving (detouring) a route using such a link reduces the occurrence of resource competition thereafter. Also,
It is expected that the number of routes with a large number of via links (long route) will be greater than the number of routes with a short route (short route), so that the possibility of a search failure at the time of detour route search is reduced.

As the second method, there is a method of determining by the detour cost. In this method, a bypass route is obtained in advance, and the bypass cost is obtained by subtracting the current route cost from the cost of the bypass route. However, if the detour route cannot be obtained, a sufficiently large value is set for the detour cost.
The detour is started from the route with the smallest detour cost.

As a result, the detour cost can be regarded as an increase in the cost due to the detour. Therefore, by detouring from a route with a low detour cost, deterioration in solution quality due to cost increase is reduced.

As a third method, there is a method of determining by winning or losing negotiation. In this method, each time negotiation is performed, the priority is decided by some means among the target agents (for example, the route is preferentially decided from the agent which receives a random number having a large value by random number generation), and the low priority is given. Perform detour processing with the received agent. In some cases, it may be effective to determine the detour order in consideration of the wins and losses of past negotiations so that the wins and losses are not overly biased.

As a result, only a specific agent wins or loses too much, so that a route with an extremely low cost cannot be created.

Furthermore, a description will be given of how to deal with the case where a request or resource is dynamically converted. Since each agent operates in asynchronous parallel, it can be seen that resource contention occurs when another agent later tries to secure a link once secured by one agent. Therefore, whether all requests are given first or requests are given sequentially in chronological order, the same mechanism can be used for solution, and it is possible to flexibly respond to dynamic changes in requirements. it can. Furthermore, when resource competition occurs, it has a mechanism for requesting an alternative plan assuming that the link does not exist, and therefore, it is possible to cope with the case where network resources dynamically change due to congestion or failure.

In the above-mentioned FIG. 3, when checking the matching of the subproblems assigned by the agents A and B, when the same edge is assigned, the negotiation C occurs respectively. In the case of task division, since the negotiation occurs at the intersection of agents, the occurrence probability decreases relatively. Furthermore, even if negotiation occurs, in the case of task division,
It can be said that it is easy to avoid because there is only one place on average. Fewer negotiations means that the partial problem solving is closed in the agent, so it is less likely to affect other agents. Therefore, it is also effective for dynamically changing the route setting request.

Since each agent operates asynchronously and independently, it can be said that the negotiation occurs when another agent tries to secure the edge after the edge once secured by one agent. Therefore, even if all route setting requests are given first,
Even if the route setting requests are given in chronological order,
The problem can be solved by the same mechanism as that of the above embodiment, and it is possible to flexibly respond to a dynamic route setting request.

Further, when a negotiation occurs, the edge is not present and a mechanism for requesting an alternative is provided. Therefore, it is possible to cope with a case where network resources dynamically change such as congestion or failure.

[0085]

As described above, according to the assignment problem-oriented cooperative problem solving system of the present invention, a problem can be solved by dividing it into a number of sub-problems according to a request and by arranging each agent in the sub-problem. It is performed locally, and by coordinating between these agents, it is possible to flexibly respond to resources such as large-scale non-regional networks, and to make requests such as route setting and to keep resources such as networks quiet. Not only when given
It is also possible to change dynamically when a capacity constraint violation occurs.

Further, according to the present invention, since each agent performs local solution calculation and resource conflict check in asynchronous parallel,
In conflict resolution, the occurrence of subsequent resource conflict can be reduced as much as possible.

[Brief description of drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a diagram for explaining the operation of the present invention.

FIG. 3 is a diagram for explaining a task division resource allocation problem.

FIG. 4 is a system configuration diagram of an embodiment of the present invention.

FIG. 5 is a diagram for explaining the outline of the operation of the embodiment of the present invention.

FIG. 6 is a diagram for explaining problem solving by each agent according to an embodiment of the present invention.

FIG. 7 is a sequence chart showing a procedure of negotiation according to an embodiment of the present invention.

FIG. 8 is a diagram for explaining a problem applied to a network resource allocation problem according to an embodiment of the present invention.

FIG. 9 is a diagram for explaining partial problem solving by an agent applied to a network resource allocation problem according to an embodiment of the present invention.

FIG. 10 is a diagram for explaining negotiation between agents applied to the network resource allocation problem of the exemplary embodiment of the present invention.

FIG. 11 is a diagram for explaining a conventional resource allocation problem.

FIG. 12 is a diagram for explaining a network resource allocation problem in which resources such as networks are allocated to requests such as route setting.

FIG. 13 is a diagram for explaining the area division type resource allocation problem.

[Explanation of symbols]

 1 Node 2 Edge 3 Request 4 Allocation Result 5 Agent 6 Problem Solving Engine 7 Resource Information 8 Knowledge of Problem Solving 9 Agent A Allocation Result 10 Agent B Allocation Result 100 Allocation Request 110 Request Distribution Means 200 Agent 220 Local Solution Calculation Means 230 Resource Information Storage Means 240 Allocation Solving Means 241 Conflict Occurrence Judging Means 242 Local Information Exchange Means 243 Request Determining Means 300 Problem Solving Knowledge 400 Resource Information 500 Allocation Results 1001 Request Distribution Processing 1002 Local Solution Calculation Processing 1003 Allocation Resource Information Storage Processing 1004 Resources Conflict check processing 1005 Local solution information exchange processing 1006 Resource conflict resolution processing 1007 Information on allocated resources 1008 End determination processing 2421 Competitive resource information exchange means 2422 Resource removal means

 ─────────────────────────────────────────────────── ─── Continued Front Page (72) Inventor Fumio Hattori 1-1-6 Uchisaiwaicho, Chiyoda-ku, Tokyo Nihon Telegraph and Telephone Corporation

Claims (3)

[Claims] 1. A system for solving an allocation problem in which a plurality of limited resources are allocated to a plurality of allocation requests, request distribution means for deciding an agent that executes allocation processing for each allocation request, and information about the resources. And problem-solving knowledge including knowledge for an agent to obtain an optimum solution to an allocation request for solving a problem and knowledge for solving a resource conflict that has occurred between the agent's solution and Resource information about the resource to be allocated, local solution calculation means for each agent to obtain a local solution satisfying the allocation request distributed by the request distribution means by using the problem solving knowledge and the resource information, and each agent Refers to the resource information storage means for storing information on resources allocated by the local solution calculation means, and For assignment problem characterized by including assignment solving means for finding a solution of resource assignment according to the constraint on the resource of the local solution obtained by each agent among all the agents and the state of competition with other agents Cooperative problem solving system. 2. The allocation solving means is a conflict occurrence determining means for checking whether or not there is a resource conflict between the local solutions calculated by the local solution calculating means by each agent, and when there is a conflict by the conflict occurrence determining means. , The local solution information exchanging means for voluntarily synchronizing and exchanging information about the local solution between the agents having the resource competition, and the resource competition is resolved from the information exchanged by the local solution information exchanging means. In order to solve the problem, the system further comprises a request determining means for determining a request to search for a local solution again, and the conflict solving system for assignment problem according to claim 1, wherein the conflicts are independently resolved among the agents in conflict with each other. . 3. The local solution information exchanging means includes competing resource information exchanging means for exchanging information on resources for which resource competition has occurred so far, and resource removing means for removing the resource from the object used for allocation. A collaborative problem solving system for assignment problems according to item 2.